Fixes and improvements to NCCL user buffer registration feature.#122
Merged
Conversation
Signed-off-by: Josh Romero <joshr@nvidia.com>
Signed-off-by: Josh Romero <joshr@nvidia.com>
Signed-off-by: Josh Romero <joshr@nvidia.com>
Signed-off-by: Josh Romero <joshr@nvidia.com>
Collaborator
Author
|
/build |
|
🚀 Build workflow triggered! View run |
|
✅ Build workflow passed! View run |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR fixes a long-standing issue with the NCCL user buffer registration feature (enabled via
CUDECOMP_ENABLE_NCCL_UBR) and makes several improvements to registration handling/cleanup.The main fix is that the calls to
ncclCommRegisterwithincudecompMallochave always incorrectly registered the wrong memory address. In particular, the registration calls were passedbuffer, which in the context ofcudecompMallocis a pointer to thee allocation pointer (e.g.,void **), but it should have been passed*buffer. Critically, this meansCUDECOMP_ENABLE_NCCL_UBRhas essentially been a no-op up to now. Users who previously tried this feature and did not see a performance impact should consider giving it another try after this patch lands.Outside this main fix, this PR also adds some code to improve registration clean up within
cudecompFreeandcudecompFinalize.